A study on subjective evaluation of facial polygon model

In order to improve drawing speed without degrading image quality in computer graphics, reduction of the number of polygons that are difficult to detect subjectively is preferable. In particular, the number of vertices of an actual facial image is huge and the reduction of these vertices without subjective degradation is strongly needed. However, the way one deletes vertices uses evaluation functions such as surface area difference, volume difference, and so on. We have developed an interactive presentation system for evaluating 3D-polygon models. The system enables an evaluator to rotate both the original and reduced polygon models in any direction simultaneously, and it gives a mean opinion score. In an experiment, 6 reduced polygon models (the number of polygons being reduced 80%, 60%, 40%, 20%, 10% and 5%) using 6 evaluation functions are evaluated on the system. From the results, the vertex removal method minimizing the change of surface area shows good scores for all the reduced models